Skip to main content

Full text of "Enhanced standoff sensing resolution using quantum illumination"

See other formats


Enhanced standoff sensing resolution using 
quantum illumination 



Saikat Guha* and Jeffrey H. Shapiro 



*Disruptive Information Processing Technologies group, Raytheon BBN Technologies, 
10 Moulton Street, Cambridge, MA 02138, USA 
^Research Laboratory of Electronics, Massachusetts Institute of Technology, 
77 Massachusetts Avenue, Cambridge, MA 02139, USA 

Abstract. Loss and noise quickly destroy quantum entanglement. Nevertheless, recent work has 
shown that a quadrature-entangled light source can reap a substantial performance advantage over 
all classical-state sources of the same average transmitter power in scenarios whose loss and noise 
makes them entanglement breaking [1, 2, 3, 4, 5, 6], standoff target-detection being an example. 
In this paper, we make a first step in extending this quantum illumination paradigm to the optical 
imaging domain, viz., to obtain better spatial resolution for standoff optical sensing. Our canonical 
imaging scenario — restricted, for simplicity, to one transverse dimension — is taken to be that of 
resolving one versus two closely-spaced in-phase specular point targets. We show that an entangled- 
state transmitter, which uses continuous-wave-pumped spontaneous parametric downconversion 
(SPDC), achieves an error-probability exponent that exceeds that of all classical-state transmitters of 
the same average power. Using these error-exponent results, we find the ultimate spatial-resolution 
limits for coherenl-slale and SPDC imaging systems that use their respective quantum-optimal 
receivers, thus quantifying the latter's spatial-resolution advantage over the former. We also propose 
a structured optical receiver that is ideally capable of harnessing 3 dB (of the full 6dB) gain in the 
error-probability exponent achievable by the SPDC transmitter and its quantum-optimal receiver. 

Keywords: Entanglement, quantum imaging, spontaneous parametric downconversion 
PACS: 42.50.-P, 42.30.-d 

Let us consider the task of discriminating between one on-axis point target vs. two 
off-axis point targets placed at angles ±0 from the on-axis direction. The transmitter 
uses a quasimonochromatic (center frequency Oorad/s), continuous- wave spontaneous 
parametric downconversion (SPDC) source, transmitting the signal beam towards the 
target region. The idler beam is retained, without loss, for subsequent joint measure- 
ment with the noisy return from the target region. Each T-sec-long transmission com- 
prises M — WT 1 signal-idler mode pairs, where W is the source's phase-matching 
bandwidth. The M signal-idler mode pairs {as^,aj^^^},l < m < M, are in independent 
identically-distributed (iid), zero-mean, jointly Gaussian states with the entangled joint 
state of each mode pair described by the Wigner-distribution covariance matrix 



As/ 



4 



1 



' S Cq 

Cq S 

5 

-C 



■ 





s 




(1) 



where Ns is the average photon number per signal (or idler) mode, S = 2Ns + 1 and Cq = 
2-\/A^(^\^^FT)- The signal beam undergoes an L-m vacuum propagation and reflects off 



the specular point target(s). The return field is incident on a length-D hard-aperture pupil. 
The roundtrip transmissivity is fc -C 1, and A^s > 1 is the mean noise photon number per 
mode. The receiver jointly detects and processes the target-return and the retained idler 
modes in order to make a minimum probability of error (MPE) decision as to which one 
of the two hypotheses (one versus two targets) is true. Only the signal-bearing spatial 
modes of the return field, i.e., (x) = \/l/D rect(jc/Z)) when there is one target present 
and ^2{x) — A{ks,d,D) cos{ksOx)xQci{x / D) when two targets are present, are relevant 
to the MPE receiver, where A:s = '^.Tc/Xs is the wave number at the signal wavelength 
Xs. Here, A{ks,d,D) = ^{1/D)/{1 + smc{ksdD)) is a normalization constant (with 
sinc(z) = sin(;rz) / 7Tz) chosen such that n, the mean photon number per mode of the field 
collected by the receiver pupil, is the same under both hypotheses. This normalization 
ensures that the receiver doesn't base its decision on a signature in the difference in 
mean photon fluxes under the two hypotheses. Because the target-return spatial modes 
are not orthogonal, we employ a Gram-Schmidt procedure to construct an orthonormal 
mode pair {^i{x),^2{x)}, with ^i{x) — ^i(jc) and 02(jc) proportional to cos{ksdx) — 
sinc{ksOD/2), which will contain all the relevant target-return information for an MPE 
decision. Under hypothesis Hi (one target present), the three modes associated with the 
return-idler annihilation operators {a(^,^^ ,a02^^ for 1 < m < M are in iid zero-mean 
jointly Gaussian states with Wigner-distribution covariance matrix 



A 



(1) 





■ Ai 





Ci 








" 







Di 














1 


Ci 





s 











4 











Ai 





Ci 
















Di 
















-Ci 





s 



(2) 



where Di = 2Nb -f 1, Ai = 2kNs + Di and Ci = y/KCq. Likewise, under hypoth- 
esis H2 (two targets present), the modes associated with the annihilation operators 
{ , dtf^2^ , a/^ } for 1 < m < M are in iid zero-mean jointly Gaussian states with Wigner- 
distribution covariance matrix 



^(2) _ 





"A2 


B2 


C2 








" 




B2 


D2 


E2 











1 


C2 


E2 


s 











4 











A2 


B2 


-C2 













B2 


D2 


-E2 













-C2 


-E2 


s 



(3) 



A2 = 2A^ kNs + Di,B2 = IABkNs, C2 = ACy , D2 = 2B^ kNs + Di and E2 = BC\ , with 



l+smc{ksdD) 



smc{ksOD/2), and B 



2smc^{kseD/2) 



1 + smc{kseD) 



(4) 



Under coherent- state illumination, the pupil-plane return field spatial modes {a/?^}, 
1 < m < M, are in non-zero-mean thermal states, with ^yphotons/m-unit mean field 
^/n ^k{x), under hypothesis ,fork= 1,2. 



(a) 



E 



I 



{«^0ifc} 

{^02fcl - PC 



'X 



Extract two 
orthogonal 
spatial modes, 
say using a pair 
of volume 
tiolograms 



(a/ J 




Sum of the 
difference 
photon counts 
^t('iO^(''0 overall M 
^ ^ modes 






^ 






: 


(M = 10"^) 


1.5x 1 \ 


K = 10"^ 




Nb = 1 




VSNR 


"'-^ 




A 

1 




1 
1 


4 



28 



7.5 8 8.5 

logio(Af) 



13 I -32 -31 -30 -29 
Ns = 0.5 lOlogio(SNR) iVs = 1.5 



FIGURE 1. (a) The phase-conjugate (PC) receiver for the one-vs.-two problem, (b) quantum Cher- 
noff bounds for the coherent-state (blue) and SPDC (red) transmitters, and probability of error for the 
SPDC-transmitter PC-receiver combination (green), versus number of modes M, (c) resolution (minimum 
resolvable angle for a given maximum error probability) vs. received-photon signal-to-noise ratio for the 
coherent-state (blue) and SPDC (red) transmitters, in each case assuming an optimal receiver 



The binary detection problem is the MPE discrimination between hypotheses Hi and 
H2 using the optimal measurement on, either, (i) the joint quantum state of the M re- 
turn modes (p^''^'*)®^ for = 1 or 2 (for the coherent-state transmitter), or (ii) the joint 
quantum state of the 3M return-idler modes, (p^^^j)'^^ for = 1 or 2 (for the quantum- 
illumination transmitter). The minimum probability of error is given by, /'^^^j^ = [1 — 
I„7i+V2, where 7,!+^ are the non-negative eigenvalues of (pW)®^ - {p(0))®M |-g]^ 
where we use p*^^^ to denote either p^^^ (coherent- state transmitter) or p^^^j^j (QI trans- 
mitter). The quantum Chernoff bound (QCB), given by 2qcb — mino<.s<i2^ where 
Qs = Tr [(p'^ ''-')■* (p *^ ^) ) ^ , is an exponentially-tight upper bound to Pj^^^^ [9]. In par- 
ticular, we have 

P^Jin - 22qcb < 2 ^0^5' (5) 

where the second inequality is the (generally looser) Bhattacharyya bound, and Qqcb = 
-limM^oc[ln(2Pe,inin)/M]. The QCB is often written as P^^^-^ < e^^^Q /2, where Rq = 
— ln(2QCB) is the error exponent. 



The error probability results are shown in Figs. 1(b) and 1(c). We have chosen = 

0. 5 . /D, and have used Ns — 0.01 mean transmitted photon-number per mode, Nb — 20 
mean noise photons per mode and the roundtrip transmissivity JC = 0.01. Even though 
we haven't calculated the QCB error exponent analytically, numerical evaluation of the 
QCB using the procedure outlined in [7] has revealed — analogous to our results for the 
target-detection problem [3] — that the SPDC-transmitter's Chernoff bound is stronger 
than the coherent-state transmitter's Chernoff bound by 6 dB (factor of 4) in the error 
exponent when Ns ^ I, k <^ I, and ^ (see Fig. 1(b)). The exponential tightness 
of the QCB then implies that a quantum illumination transmitter has a factor-of-four 
advantage in —\n(Pe^J over a coherent- state transmitter when both radiate the same 
average photon number per mode and both use optimum receivers. Figure 1(c) plots the 
minimum resolvable target-separation angle as a function of the received signal-to-noise 
ratio (SNR) kNs/Nb, for a Pe = 0.03 probability of error threshold. The resolution- 
versus-SNR plots show the typical divergence behavior: below a threshold SNR the two 
targets cannot be resolved. Quantum illumination's error-exponent advantage enables it 
to have an appreciably lower threshold SNR than does the coherent- state transmitter. 
Thus, for the parameter values chosen (k = 10^^ and A^^ = 1), the resolution plots 
represent almost a 2-dB lateral shift along the SNR axis. For higher A^^ values, quantum 
illumination's SNR advantage is more pronounced. 

What these results do not tell us yet, is a structured realization of the receiver (in 
conjunction with the SPDC transmitter) that would be able to harness this 6dB error 
exponent advantage over the coherent- state transmitter. Figure 1(a) shows a structured 
receiver that can bridge half this gap [5], i.e., obtain an error-probability exponent 3 dB 
higher than the coherent state transmitter's optimal error exponent. The phase-conjugate 
(PC) receiver conjugates the return modes a^j^ and mixes them with corresponding 
idler modes dj^ on a 50-50 beam splitter. Balanced detection of the resulting ouputs, 
as shown in Fig. 1(a), and summing over all M modes, then provides a sufficient statistic 
for deciding between the two hypotheses. Quantum illumination with the PC receiver 
can be shown, analytically, to enjoy a 3 dB error-exponent advantage over the coherent- 
state system when Ns<^1,Nb^I,k<^1, and 0^0. Extending quantum illumination 
to more complex imaging scenarios, and building a structured design of the quantum- 
optimal receiver are subjects of ongoing investigation. 

This work was supported by the DARPA Quantum Sensors Program. 

REFERENCES 

1. M. F. Sacchi, Phys. Rev. A 71, 062340 (2005); M. F. Sacchi, Phys. Rev. A 72, 014305 (2005). 

2. S. Lloyd, Science 321, 1463 (2008). 

3. S.-H. Tan, et. al, Phys. Rev. Lett. 101, 253601 (2008). 

4. S. Guha, Proc. of IEEE Int. Symp. on Inf. Th. (ISIT), Seoul, South Korea (IEEE, New York), (2009). 

5. S. Guha and B. I. Erkmen, Phys. Rev. A 80, 052310 (2009). 

6. J. H. Shapiro, Phys. Rev. A 80, 022320 (2009). 

7. S. Pirandola and S. Lloyd, Phys. Rev. A 78, 012331 (2008). 

8. C. W. Helstrom, Quantum Det. and Est. Theory, Vol. 123, Academic Press, New York (1976). 

9. K. M. R. Audenaert, et al, Phys. Rev. Lett. 98, 160501 (2007); J. CalsamigUa, et at., Phys. Rev. A 
77, 012331 (2008).